.
F U L L T E X T S O U R C E : BioRxiv
Understanding the expression changes that come with age is an important step in understanding the ageing process as a whole. By combining such transcriptomic data with other sources of information, for instance protein-protein interaction (PPI) data, it is possible to make inferences about the functional changes that occur with age. To address this, we conducted a meta-analysis on 127 publicly available microarray and RNA-Seq datasets from mice, rats and humans, to identify genes that are commonly differentially expressed with age in mammals. We also conducted analyses on subsets of these datasets, to produce transcriptomic signatures for brain, heart and muscle tissues, all of which are important tissues in the pathophysiology of ageing. This approach identified the transcriptomic signatures of the ageing system, as well as brain, heart and muscle tissues. We then applied enrichment analysis and machine learning to functionally describe those signatures. This revealed a typical ageing signature including the overexpression of immune and stress response genes and the underexpression of metabolic and developmental genes. Further analysis of the ageing expression signatures revealed that genes differentially expressed with age tend to be broadly expressed across tissues, rather than be tissue-specific, and that the ageing expression signatures (particularly the overexpressed signatures) of the whole system, brain and muscle tend to include genes that are central in PPI networks. We also show that genes underexpressed in the brain are highly central in a co-expression map, suggesting that underexpression of these genes may play a part in cognitive ageing. In sum, we show numerous functional similarities between the ageing transcriptomes of these important tissues, a broad non-specific expression pattern in genes differentially expressed with age, along with altered network properties of these genes in both a PPI and co-expression network.
The expression signature of ageing, i.e. the characteristic changes in gene expression that can be expected in an ageing organism, is of vital importance to understanding the ageing process and how interventions could modulate it. Knowledge of expression patterns in ageing organisms can be employed as biomarker panels that estimate a ‘transcriptomic age’ (Peters et al., 2015), in addition to giving insight into the basic processes associated with ageing (Stegeman and Weake, 2017) and serving as a starting point from which to identify drugs and other interventions that may assist with healthy ageing (de Magalhaes et al., 2012).
Comparative analysis of gene expression data across species is a powerful method to determine an expression signature of ageing. Previously meta-analyses of gene expression with age in mammals have identified changes in stress responses, metabolism and immune response genes (de Magalhães, Curado and Church, 2009) while meta-analysis of the dietary restriction expression signature identified novel changes in retinol metabolism and copper-ion detoxification in this important ageing modulating process (Plank et al., 2012).
Modern techniques allow such analyses to be taken further. Machine learning is increasingly being used in ageing research and offers a lot of potential for the identification of ageing and ageing-related disease genes (Fabris, De Magalhães and Freitas, 2017). Machine learning methods complement traditional bioinformatics analyses, providing a different perspective and with the potential for more predictive results (Fabris et al., 2019). Because of this, machine learning offers promising methods by which to identify genes of interest and potential intervention targets for the alleviation of ageing and ageing-related diseases.
Here, we have performed a meta-analysis of ageing using the methods of de Magalhães, et al. (2009) on 127 microarray and RNA-Seq datasets from humans, mice and rats, and applied machine learning alongside enrichment methods to analyse the results. This revealed an ageing signature consistent with previous analyses. In addition, we performed analyses on tissue-specific subsections of these datasets for brain, heart and muscle which revealed some interesting tissue specific differences in connectivity.
.../...
4 Results
The global meta-analysis identified 449 genes overexpressed with age and 162 underexpressed with age. This is considerably more than the results of de Magalhães, et al. (2009), where 56 overexpressed and 17 underexpressed genes were identified. For the tissue-specific analyses, in brain 147 genes were overexpressed and 16 genes were underexpressed, in heart 35 genes were overexpressed and 5 genes were underexpressed, and in muscle 49 genes were overexpressed with 73 genes underexpressed. The top-5 overexpressed genes for each analysis are presented in Table 1 and the top-5 underexpressed genes for each analysis are presented in Table 2.
Table 1.
Top-5 genes most consistently overexpressed with age across datasets for all tissues and for each tissue studied. The value given between brackets in the ‘p-value’ column header is the p-value threshold at which FDR <0.05.
Table 2.
Top-5 genes most consistently underexpressed with age across datasets for all tissues and for each tissue studied. The value given between brackets in the ‘p-value’ column header is the p-value threshold at which FDR <0.05.
As with de Magalhães, et al. (2009), the most significantly overexpressed genes in this meta-analysis were principally involved in immune responses and inflammation, in particular in the global and the brain-specific analyses. Several complement proteins were overexpressed in these analyses, with C1QA appearing at the top of both the global and brain-specific analyses, C1QC likewise appears in both top-5 lists. The top genes in the heart-specific results include the structural protein gene MGP, genes involved in amine metabolism and oxidation-reduction processes (MAOA and VAT1) as well as the iron and copper metabolism gene CP. In muscle the top overexpressed gene was CDKN1A, a cell cycle regulator. Other interesting top genes overexpressed in muscle include EFEMP1, a gene involved in eye morphogenesis and CHRNA1 that codes for a muscle acetylcholine receptor subunit.
A common theme across the top underexpressed genes is mitochondrial metabolism. In the global results, the top underexpressed gene is UQCRFS1, a subunit of mitochondrial complex III, while in heart NDUFS7, a component of mitochondrial complex I, is the second most significantly underexpressed gene. Another mitochondrial complex I subunit, NDUFC1 was the third most significantly underexpressed gene in muscle. The brain is the only tissue studied that did not see an underexpression of mitochondrial genes. Indeed, all the top-5 genes underexpressed in the brain signature have clear roles in neuronal signalling and/or development. Complete lists of all significant genes for all the analyses can be found in Supplementary Tables S2-S9.
Interestingly, several genes with known involvement in ageing-modulating pathways were differentially expressed for instance IGF1 was overexpressed, while IGF2R and RICTOR were underexpressed in the tissue non-specific analysis.
.../...
F O R T H E R E S T O F T H E S T U D Y, P L E A S E V I S I T T H E S O U R C E .
.
